【作者】 李兵；

【导师】 连小珉；

【作者基本信息】 清华大学 ， 机械工程， 2010， 博士

【摘要】 目前的声学摄像机系统,能够直观、快速地进行噪声源的识别和定位,但由于运动汽车噪声的多普勒效应等问题,使得上述系统在运动汽车声场可视化中的应用具有一定的难度,并且现有系统在应用于运动汽车时常包括一些不方便或不够准确的操作,影响了测量的效率。因此,研究能够直观、快速、准确地实现运动汽车声场可视化的方法,就具有非常现实的意义。本文首先建立起运动汽车声场可视化中各工程对象的数学描述。建立描述图像和空间中的位置的各类坐标系和不同坐标间的变换关系,定义了声源场、测量场、重建场等主要声场,并描述了声场间的时空关系。提出了带大地坐标位置的汽车图像的双目视觉获取方法。通过两台摄像机分别拍摄运动汽车的图像,识别汽车表面粘贴的定位标志在图像中的位置,通过编号识别匹配相同的位置点,进而利用双目视觉原理计算其空间位置,得到汽车在运动过程中的位置和速度。提出了运动声场分析的虚假声源抑制方法。通过对短时多普勒效应的分析,建立运动声场的短时波叠加关系;利用波束形成方法对声源场内的声源点的强度进行预估,根据预估结果得到可能的主要声源点;针对各声源点建立短时波叠加方程,并求解声源强度;利用求解的声源强度计算重建场声压分布。通过摄像机触发信号与声压信号的同步采集和对触发信号的分析,实现声场与汽车图像的时间同步;通过对重建点的大地坐标位置和其对应的屏幕位置的变换实现声场与汽车图像的空间匹配;通过声场色标映射实现声场与汽车图像的融合显示,并最终合成动态声场视频。最后,搭建了应用上述方法进行运动汽车声场可视化的软、硬件试验平台,并提出试验的操作流程,通过对已知声源和实车噪声声源的可视化测量试验对系统性能进行验证。试验结果表明,运动汽车声场可视化系统可以准确、方便地进行运动汽车的声场可视化重现,从而能够直观地从结果视频中看到的主要噪声源,并且能够看到声场在汽车运动过程中的变化情况。更多还原

【Abstract】 With current acoustic camera systems, noise sources can be identified and located quickly and intuitively. But because of the Doppler Effect of the noise of moving vehicles, it’s difficult to apply these systems directly in the sound visualization of moving vehicles. Besides, there are some inconvenient or inaccurate procedures when these systems used on moving vehicles, which make the measurement inefficient. Therefore, the researches on the method that makes sound visualization of moving vehicles more intuitively, quickly, and accurately have very practical significance.First, the mathematical descriptions are created for all kinds of the engineering objects in the sound visualization of moving vehicles, including the coordinate systems for the graphic and spatial postions, and their transformations. The sound fields are defined, such as the source field, the measurement field, and the reconstruction field, and their temporal and spatial relationships are desribed.The binocular vision method for acquiring the vehicle pictures with the ground positions is proposed. The pictures of the moving vehicle are shot by two cameras separately, and the screen positions of the location marks stuck on the side surface of the vehicle are identified. The same marks are matched with identifying the serial numbers of the marks, and the spatial positions of the marks can be computed with binocular vision theory, then the position and the velocity of the moving vehicle are acquired.The method of restraining the ghost sources for the moving source sound field is proposed. The short-time wave superposition relation for moving sound sources is deduced by analyzing the Doppler Effect in very short time. The source strengths of the point sources in the source field are pre-estimated with the beamforming algorithm, and the possible main sources are identified with the pre-estimation result. Then the wave superposition equation for the main sources is founded, and the source strengths are figured out by solving the equation. At last, the sound pressure distribution of the reconstruction field is computed with the source strengths.The sound field and the pictures are temporally synchronized by collecting the sound pressure and the camera trigger signal at same time and analyzing the saved trigger signal. The sound field and the pictures are spatially matched with the coordinate transformation between the ground coordinates and the screen coordinates of the reconstruction points. Then the sound field and the pictures are merged with the mapping from sound pressure to color, and the dynamic sound field video is finally generated.The hardware and software system of the sound visualization of moving vehicles with the above method is developed, and the procedure of the test is established. The visualization tests with known sound sources and real vehicle noises are carried out to verify the performance of the system. The results of the tests indicate that the sound visualization of moving vehicles can be realized correctly and conveniently with this system. Then, the main noise sources can be identified with the analysis to the reconstruction video intuitively, and the changes of the sound field in the moving process can be represented.更多还原